The bias towards addition rather than subtraction

why we prefer to add rather than to subtract or modify

All codified and legible systems (software, laws, regulations, organisations) become more complex over time because although we constantly add new features to these systems, we never take the time and effort to subtract them. In fact, we prefer to add new components rather than changing existing components.

There is evidence that our preference for addition rather than subtraction is a fundamental cognitive bias. However, a preference for additive changes is not just an irrational preference.

Chesterton’s Fence is an inadequate defence against the risks of subtraction

The principle of Chesterton’s Fence can be summarised as follows: Don’t remove any component/feature of the system if you do not understand its function and why it exists. Subtracting is a much riskier activity than adding as you may inadvertently remove critical system functionality.

But even this principle is inadequate in protecting us from the risks of subtraction. In most complex systems, degeneracy i.e. the existence of multi-functional components with partially overlapping functions, is the norm rather than the exception.

The bias towards addition in software

This risk is most apparent in software systems where programmers have good reason to prefer to add code rather than to subtract code. This is true even when the codebase has good test coverage - every software system is almost certainly being used in the wild in a manner that was not anticipated at the time of its design. Programmers even prefer to add new modules than risk changing a common module, as changing a common module risks propagating changes throughout the system in an unpredictable manner.

The bias towards addition in laws and regulations

New laws and regulations typically take the form of “political kludges” i.e. “piecemeal modifications that patch over old programs rather than replace them”. There are plenty of political reasons for this but the overwhelming reason for it is entirely logical and rational:

interdependencies between rules create entanglements that hinder the undoing of existing policy. Deletion of a rule may cripple other dependent rules that rely on the functionality of the deleted rule…..

Deletion has the benefit of reducing complexity. However, the deletion process is stymied by entanglement. Unfavorable rules cannot be deleted surgically; they may be entangled with favorable rules that have to be deleted as well.

As an example,

Consider the Alternative Minimum Tax (AMT) in the US Tax Code. Many observers deem the AMT to be an unsatisfactory solution to the problems it was intended to solve, but also believe that it will be difficult to undo or significantly edit the AMT because many other aspects of the federal tax system have come to rely on the AMT.

The bias towards addition explains the explosion in the scope and scale of laws and regulations worldwide. For example,

In America, the number of federal regulations has more than doubled since 1970. The total word count of Germany’s laws is 60% larger today than it was in the mid-1990s.

But it also explains why every long-lived codified system, private or government, suffers from the symptoms of not just software bloat but complexity bloat. Even a successful firm such as Google, sooner or later, becomes “terminally bureaucratic”.

The risks of subtracting components in a complex system are why reforming or re-engineering existing codified systems is an uphill task and why you should never attempt a system rewrite or even a major system overhaul.

Don’t scar on the first cut

So what can be done? The first step is to stop doing more harm and a useful heuristic is not to be so quick to add rules, regulations and “safety clutter” every time something goes wrong. As  David Heinemeier Hansson puts it,

Policies are often the result of something that once went wrong. It’s organizational scar tissue developed from a This Can Never Happen Again mandate. And its almost always ill-considered.

The problem with policies are that they compound and eventually add up to the rigidity of bureaucracy that everyone says they despise.